Examples with CoNLLWord com.hankcs.hanlp.corpus.dependency.CoNll.CoNLLWord used on opensource projects

Example 1 with CoNLLWord

use of com.hankcs.hanlp.corpus.dependency.CoNll.CoNLLWord in project HanLP by hankcs.

the class MaxEntDependencyModelMaker method generateUniContext.

public static Collection<String> generateUniContext(CoNLLWord[] word, int i, int j) {
    Collection<String> context = new LinkedList<String>();
    context.add(word[i].NAME + '→' + word[j].NAME);
    context.add(word[i].POSTAG + '→' + word[j].POSTAG);
    context.add(word[i].NAME + '→' + word[j].NAME + (i - j));
    context.add(word[i].POSTAG + '→' + word[j].POSTAG + (i - j));
    CoNLLWord wordBeforeI = i - 1 >= 0 ? word[i - 1] : CoNLLWord.NULL;
    CoNLLWord wordBeforeJ = j - 1 >= 0 ? word[j - 1] : CoNLLWord.NULL;
    context.add(wordBeforeI.NAME + '@' + word[i].NAME + '→' + word[j].NAME);
    context.add(word[i].NAME + '→' + wordBeforeJ.NAME + '@' + word[j].NAME);
    context.add(wordBeforeI.POSTAG + '@' + word[i].POSTAG + '→' + word[j].POSTAG);
    context.add(word[i].POSTAG + '→' + wordBeforeJ.POSTAG + '@' + word[j].POSTAG);
    return context;
}

Example 2 with CoNLLWord

use of com.hankcs.hanlp.corpus.dependency.CoNll.CoNLLWord in project HanLP by hankcs.

the class MaxEntDependencyModelMaker method generateSingleWordContext.

public static Collection<String> generateSingleWordContext(CoNLLWord[] word, int index, String mark) {
    Collection<String> context = new LinkedList<String>();
    for (int i = index - 2; i < index + 2 + 1; ++i) {
        CoNLLWord w = i >= 0 && i < word.length ? word[i] : CoNLLWord.NULL;
        // 在尾巴上做个标记，不然特征冲突了
        context.add(w.NAME + mark + (i - index));
        context.add(w.POSTAG + mark + (i - index));
    }
    return context;
}

Example 3 with CoNLLWord

use of com.hankcs.hanlp.corpus.dependency.CoNll.CoNLLWord in project HanLP by hankcs.

the class TestParse method testEvaluate.

public void testEvaluate() throws Exception {
    testParse();
    LinkedList<CoNLLSentence> sentenceList = CoNLLLoader.loadSentenceList("D:\\Doc\\语料库\\依存分析训练数据\\THU\\dev.conll");
    Evaluator evaluator = new Evaluator();
    int id = 1;
    for (CoNLLSentence sentence : sentenceList) {
        System.out.printf("%d / %d...", id++, sentenceList.size());
        long start = System.currentTimeMillis();
        List<Term> termList = new LinkedList<Term>();
        for (CoNLLWord word : sentence.word) {
            termList.add(new Term(word.LEMMA, Nature.valueOf(word.POSTAG)));
        }
        CoNLLSentence out = CRFDependencyParser.compute(termList);
        evaluator.e(sentence, out);
        System.out.println("done in " + (System.currentTimeMillis() - start) + " ms.");
    }
    System.out.println(evaluator);
}

Example 4 with CoNLLWord

use of com.hankcs.hanlp.corpus.dependency.CoNll.CoNLLWord in project HanLP by hankcs.

the class AbstractDependencyParser method parse.

@Override
public CoNLLSentence parse(String sentence) {
    assert sentence != null;
    CoNLLSentence output = parse(segment.seg(sentence.toCharArray()));
    if (enableDeprelTranslater && deprelTranslater != null) {
        for (CoNLLWord word : output) {
            String translatedDeprel = deprelTranslater.get(word.DEPREL);
            word.DEPREL = translatedDeprel;
        }
    }
    return output;
}

Example 5 with CoNLLWord

use of com.hankcs.hanlp.corpus.dependency.CoNll.CoNLLWord in project HanLP by hankcs.

the class CRFDependencyParser method parse.

@Override
public CoNLLSentence parse(List<Term> termList) {
    Table table = new Table();
    table.v = new String[termList.size()][4];
    Iterator<Term> iterator = termList.iterator();
    for (String[] line : table.v) {
        Term term = iterator.next();
        line[0] = term.word;
        line[2] = POSUtil.compilePOS(term.nature);
        line[1] = line[2].substring(0, 1);
    }
    crfModel.tag(table);
    if (HanLP.Config.DEBUG) {
        System.out.println(table);
    }
    CoNLLWord[] coNLLWordArray = new CoNLLWord[table.size()];
    for (int i = 0; i < coNLLWordArray.length; i++) {
        coNLLWordArray[i] = new CoNLLWord(i + 1, table.v[i][0], table.v[i][2], table.v[i][1]);
    }
    int i = 0;
    for (String[] line : table.v) {
        CRFModelForDependency.DTag dTag = new CRFModelForDependency.DTag(line[3]);
        if (dTag.pos.endsWith("ROOT")) {
            coNLLWordArray[i].HEAD = CoNLLWord.ROOT;
        } else {
            int index = convertOffset2Index(dTag, table, i);
            if (index == -1)
                coNLLWordArray[i].HEAD = CoNLLWord.NULL;
            else
                coNLLWordArray[i].HEAD = coNLLWordArray[index];
        }
        ++i;
    }
    for (i = 0; i < coNLLWordArray.length; i++) {
        coNLLWordArray[i].DEPREL = BigramDependencyModel.get(coNLLWordArray[i].NAME, coNLLWordArray[i].POSTAG, coNLLWordArray[i].HEAD.NAME, coNLLWordArray[i].HEAD.POSTAG);
    }
    return new CoNLLSentence(coNLLWordArray);
}